Investors, just like animals, make their best bets based on the information available to them and their previous experience. Working on quantitative neuroscience and animal behaviour for almost 10 years has given me a rich view on decision making.In my personal view, animals collect information from their senses, directly or by interacting with other individuals, and make consequent decisions. These decisions may allow them to eat, mate, hide, catch prey or simply survive another day, just like investors.
Animals make their best bets based on the available information, their instincts, and previous knowledge. Their instincts, encoded in their genome, constitute a general idea of how the world works, what specific sensory inputs may mean and the consequences of a given action. Their previous knowledge may fine-tune their instincts. This allows animals to review nature’s rules within each generation. Finally, current sensory info is plugged-in and a decision is made.
The whole idea seems flawless at first. However, as always, reality is more complicated than it seems. There’s a chance that your instincts are completely wrong, so your species survives with an idea of the world that doesn’t fit reality.There’s also a chance that your instincts were correct but your environment changes too fast and you die before you learn how to act in the new conditions. You could also be a fast learner with the right instincts, but the sensory information you collect is just not enough to make a precise enough guess of the external world.The most dramatic case happens when your instincts, your knowledge and the info you receive is complete and accurate, but the consequences of your actions are just not deterministic or are chaotic, even. This means that the same action may have different outcomes. Therefore, the task becomes to make a decision based on incomplete and potentially misleading information that may or may not give us an insight into the future consequences of our actions. (I guess the problem is starting to look familiar to anybody working in finance).
How do animals solve this puzzle?
Well, they don’t always make it. But when they do they change, they adapt, they evolve. A species can either assume that the world is too complicated and create a large number of similar offspring and hope that many of them survive. This strategy is frequently associated with very short generational lifespans. This strategy gives the option to adapt dynamically to the environment through instincts. Other species choose to evolve fewer, smarter individuals with the hope that they will learn how the world works through personal experience or interactions and then adapt their behaviour to each situation. Sensory systems are also being improved constantly; one of my favourite examples of this is the eye: believed to have evolved independently at least four times in different animal families.
So how can you solve this puzzle?
First, I would add that you want to solve this puzzle in your lifetime. Waiting for evolution to solve it for you is going to take a little while. Humans have evolved to survive in the savannah, thus our intuitions may work there, but we have not evolved to estimate financial parameters. Yet. What we have evolved to do, though, is build tools that help us. For thousands of years, we have created tools that have improved our physical capabilities-we can fly! Since the invention and development of machine learning, we can also improve our cognitive capabilities. That is exactly the kind of help we need here.
Machine learning algorithms are able to learn complex descriptions of the world based on previous information, what I previously called instinct and knowledge for animals. Perhaps more importantly, they can also collect current information and decide what to do with it. Just what we needed. Machine learning can dramatically increase our brain’s computational power. Our job now is to design the algorithm that can learn how to solve our problem. Unless you can choose all the parameters of your implementation using a pen and a paper—a closed form solution—you will need to iterate, just like evolution. Some of these algorithms are even capable of adapting to new conditions and tune their parameters to integrate the most recent information, in the same way, knowledge works for us. As the reader may have guessed by now, it isn’t by chance scientist based the design of artificial neural networks on real brains.
Putting it all together
Machine learning allows us to integrate information from different sources and temporal windows, therefore making our description of the world as simple or as complex as we need it to be. It also allows us to build this description automatically based on current conditions with a historical reference.
All of this ultimately allows us to enhance our brain capabilities to implement a new rigorous, informed and unbiased decision-making process.